(Table S1) Radiocarbon data of sediment core SO161/5 22SL

At the end of the Last Glacial Maximum (19,000 to 11,000 years ago), atmospheric carbon dioxide concentrations rose while the Delta14C of atmospheric carbon dioxide declined**1, 2. These changes have been attributed to an injection of carbon dioxide with low radiocarbon activity from an oceanic abyssal reservoir that was isolated from the atmosphere for several thousand years before deglaciation**3. The current understanding points to the Southern Ocean as the main area of exchange between these reservoirs4. Intermediate water formed in the Southern Ocean surrounding Antarctica would have then carried the old carbon dioxide signature to the lower-latitude oceans**5, 6. Here we reconstruct the Delta14C signature of Antarctic Intermediate Water off the coast of Chile for the past 20,000 years, using paired 14C ages of benthic and planktonic foraminifera. In contrast to the above scenario, we find that the delta14C signature of the Antarctic Intermediate Water closely matches the modelled surface ocean Delta14C, precluding the influence of an old carbon source. We suggest that if the abyssal ocean is indeed the source of the radiocarbon-depleted carbon dioxide, an alternative path for the mixing and propagation of its carbon dioxide may be required to explain the observed changes in atmospheric carbon dioxide concentration and radiocarbon activity.

(Table S1) Age determination of ODP Site 165-1002

A series of 14C measurements in Ocean Drilling Program cores from the tropical Cariaco Basin, which have been correlated to the annual-layer counted chronology for the Greenland Ice Sheet Project 2 (GISP2) ice core, provides a high-resolution calibration of the radiocarbon time scale back to 50,000 years before the present. Independent radiometric dating of events correlated to GISP2 suggests that the calibration is accurate. Reconstructed 14C activities varied substantially during the last glacial period, including sharp peaks synchronous with the Laschamp and Mono Lake geomagnetic field intensity minimal and cosmogenic nuclide peaks in ice cores and marine sediments. Simulations with a geochemical box model suggest that much of the variability can be explained by geomagnetically modulated changes in 14C production rate together with plausible changes in deep-ocean ventilation and the global carbon cycle during glaciation.

Stable oxygen isotope and Mg/Ca ratios of planktonic foraminifera from the Caribbean and North Atlantic

Variations in the strength of the North Atlantic Ocean thermohaline circulation have been linked to rapid climate changes during the last glacial cycle through oscillations in North Atlantic Deep Water formation and northward oceanic heat flux. The strength of the thermohaline circulation depends on the supply of warm, salty water to the North Atlantic, which, after losing heat to the atmosphere, produces the dense water masses that sink to great depths and circulate back south. Here we analyse two Caribbean Sea sediment cores, combining Mg/Ca palaeothermometry with measurements of oxygen isotopes in foraminiferal calcite in order to reconstruct tropical Atlantic surface salinity during the last glacial cycle. We find that Caribbean salinity oscillated between saltier conditions during the cold oxygen isotope stages 2, 4 and 6, and lower salinities during the warm stages 3 and 5, covarying with the strength of North Atlantic Deep Water formation. At the initiation of the Bølling/Allerød warm interval, Caribbean surface salinity decreased abruptly, suggesting that the advection of salty tropical waters into the North Atlantic amplified thermohaline circulation and contributed to high-latitude warming.

(Table 1) Age determination of surface sediment samples from the North Atlantic

Most seafloor sediments are dated with radiocarbon, and the sediment is assumed to be zero-age (modern) when the signal of atmospheric testing of nuclear weapons is present (Fraction modern (Fm) > 1). Using a simple mass balance, we show that even with Fm > 1, half of the planktonic foraminifera at the seafloor can be centuries old, because of bioturbation. This calculation, and data from four core sites in the western North Atlantic indicate that, first, during some part of the Little Ice Age (LIA) there may have been more Antarctic Bottom Water than today in the deep western North Atlantic. Alternatively, bioturbation may have introduced much older benthic foraminifera into surface sediments. Second, paleo-based warming of Sargasso Sea surface waters since the LIA must lag the actual warming because of bioturbation of older and colder foraminifera.